This is a simple Python AI Chatbot with simulated emotions. She starts out not knowing much, but learns more from talking with you.
This chatbot starts with memory files that could be edited or replaced with files of your own choosing. As you talk to her, she learns new words marked under specific emotions, along with learning new replies from you, both of which update her memory. In order to exit the conversation and allow saving of memory, simply enter "//exit" (Without the quotes). Else, the memory from that conversation won't be saved. The conversation also outputs to 2 chatlog files. One is just the regular chat, while the extended chatlog (Xchatlog) contains the conversations plus what OPHELIA is "thinking".
The only required outside program is espeak. If you choose to not use this feature, simply comment out line 100.
Update Version 1.01: OPHELIA now can recognize single terms from a user message, but only after OPHELIA has collected over 2000 words OPHELIA's emotion dictionary, and at least 500 learned reponses. Of course, if you choose to change this for your own purposes, it is easy, but it is recommended to only let that feature be activated after OPHELIA has learned a lot more. The reason is that if it is activated from the start, OPHELIA will not learn as many new responses, because terms from the user message will be detected instead, and will not activate her response learning code. Of course, this really opens the door for including even more, similar features, to be activated when OPHELIA learns even more.
Update, Experimental Script (05-10-2023): Set up a script derived from OPHELIA in an attempt to grow OPHELIA's emotion dictionary by inputting a text file and letting OPHELIA break it apart in the same way as user messages, with the results shown in the 'Dictionary_Fill_Experiment1' directory of this project repo. The initial experiment with the provided text file 'texttest1.txt' was successful, but then I aimed a bit too high. I got a text file copy of 'A Tale Of Two Cities' (Not included in the directory due to Copyright issues and other legal issues, etc.). OPHELIA picked it apart in less than 2 seconds, with OPHELIA's emotion dictionary then contianing over 13,000 words. Seemed great, but unfortunately due to the lacking dictionary OPHELIA had before that, OPHELIA ended up marking off most words as 'happy' words. It seems that OPHELIA must be taught a lot more either manually, through chat, or through simpler books first.
Update Version 1.03 (05-11-2023): Modified the code for detecting 'single word' message pieces, to now include a slightly 'random' access to it, the odds of it landing on there depends on how many emotion words and learned responses OPHELIA has in memory. The reason for this is if it was activated 100% given a certain number of emotion words and responses in memory, this would greatly impact the amount of new responses OPHELIA could learn, since it wouldn't get to the last, learning part of OPHELIA's message detection code. It seems much better to give OPHELIA more room to learn before activating that part of the code. This also means OPHELIA in some sense, can 'evolve', or 'level up' over time.
Update 'DoubleChat' (05-20-2023): Made an identical copy of OPHELIA names 'Phoebe'. The purpose was to set up a new script to allow OPHELIA to chat with Phoebe, which was successful. The script, along with both chatbots, are in the directory labeled 'DoubleChat'.